Smarsh used its EDGE event in New York to recast archived employee communications as a source of business intelligence rather than a regulatory burden. Artificial intelligence can now classify a year of email, messages and AI-prompt traffic and turn it into client context for the next meeting. Only 26% of enterprises say governance frameworks are keeping pace with AI deployment, according to a Smarsh study, which is why the company argues governance must be built into the foundation.
How Smarsh turns archived messages into context
Smarsh helps customers extract insight from communications long considered too unstructured to use, said Chief Strategy Officer Goutam Nadella. A CRM system holds only what someone recorded about a meeting, while AI can analyze conversations across email, WhatsApp, Teams and Zoom and suggest preparation steps. The company was a launch partner for compliance application programming interfaces from Anthropic and OpenAI, capturing prompts so risky outputs can be traced to sources. Customers are also building their own agents outside compliance platforms and connecting data through Model Context Protocol servers.
Supervisors who once ran complex Boolean searches for market manipulation can now ask an agent questions in plain language, said Chief Customer Officer Rohit Khanna. The system returns flagged conversations with risk scores, supported by a context layer that maps identity hierarchies and for which Smarsh has filed a patent. That layer is designed to reduce false positives by tying messages to roles and relationships. Smarsh also builds compliance language models that cost less to run than large language models, including a multilingual model that tracks speakers switching between languages.
The shift comes after billions of dollars in regulatory fines imposed on financial institutions for failing to preserve business communications on WhatsApp and other messaging platforms. Regulations remain technology-agnostic, so firms stay liable for actions of their agents, including agents communicating with other agents. AI in financial services has moved beyond pilots, yet most governance frameworks lag behind deployment. Through a multiyear strategic collaboration agreement with Amazon Web Services, Smarsh is building governance and security into the infrastructure layer on AWS, with implementation support from Caylent.
What governed communications AI means for business
For companies in regulated industries, the change turns supervision from search work into review of scored, explainable results. Compliance teams can query communications directly instead of maintaining Boolean rules, while relationship managers gain preparation briefs drawn from a full year of history. Because 70% to 80% of compliance spending still goes toward labor, the return from automation is potentially large. Smaller firms gain access to capabilities once requiring large surveillance teams, while large firms can standardize controls and data access across channels.
The same architecture imposes clear limits that buyers should verify before deployment. Smarsh describes four levels of autonomy — assisting, recommending, executing under guardrails and fully autonomous — matched to each process by reversibility, complexity and observability. Its model risk management agent, built by an 11-person volunteer S-Combinator group, is only in limited availability. Firms should ask how prompts are captured, who or what can access data, how identity mapping works, and how each agent decision is explained and audited.
The marker to watch is whether Smarsh moves its agents and MCP server from limited availability into broad production use on AWS. Wider rollout with documented audit trails and patent-backed identity context would confirm that governance can scale with autonomy. If adoption stalls at pilots, the gap between AI deployment and the 26% governance readiness figure will remain the main constraint.
